Translating documentaries. Does the integration of a bilingual glossary of domain-specific terminology into the translation process reduce the translators' workload?
5 July 2017
UGent -Vakgroep Vertalers, Tolken en Communicatie - Abdisstraat 1 - lokaal A104 - 9000 Gent
10:00 AM - 1:00 PM
Organization / co-organization:
Uiniversiteit Gent - Vakgroep Vertalers, Tolken en Communicatie / Universteit Antwerpen - Faculteit Letteren en Wijsbegeerte
Prof Véronique Hoste, Prof Aline Remael
Phd defence Sabien Hanoulle - Faculty of Arts
The thesis under study is a contribution to the research in translation of documentaries and more particularly, the translation of domain-specific terminology, one of the challenges in this field. Carrying out translation experiments, this dissertation investigates whether or not bilingual glossaries, manually or automatically extracted, reduce the workload of documentary translators.
The research consists of three major parts, all thoroughly analysed against the state-of-the-art studies. The first part concerns the analysis and the selection of the corpus. The Flemish public broadcaster VRT made a corpus of English documentaries and their Dutch translation available. In a preparatory stage, an in-depth analysis of the text type showed the corpus contained domain-specific terminology, especially in documentaries meant for informative purposes. As a consequence, an experimental corpus of three informative documentaries was selected for the translation experiments.
The second part focuses on manual and automatic terminology extraction, the underlying software of automatic term extractors and the testing of three existing systems. In order to understand the test results, two key features for terminology (termhood and unithood) were discussed and an overview of the different strategies term extractors use was provided. Annotators manually labelled all the terminology of the experimental corpus, drawing up in this way a gold standard as an objective means for testing the automatic systems. The accuracy of these systems was expressed in terms of precision and recall. The best performing system was selected to extract the glossaries for the experiments.
The third part deals with the translation experiments. In a pilot project, Master’s students in translation translated three texts first without, then with the manual or the automatic glossaries at their disposal, while a keystroke logging software registered the process. For the main experiment with professional translators, the experimental set-up was slightly modified, introducing some remedial measures learned from the pilot project. Statistical analyses of the total process time and the pause time before each term were elaborated.
The results revealed that in most working conditions the candidates worked significantly faster when translating with a glossary and that they made less terminological errors. Furthermore, the dissertation proposes an ecologically valid experimental design, tested and remediated in the ongoing research. Yet, there was room for improvement for the automatically extracted glossaries due to the small corpus and the free translation style, typical for translating documentaries.
Contact email: firstname.lastname@example.org